deep-river 0.3.2__tar.gz → 0.4.0__tar.gz

{deep_river-0.3.2 → deep_river-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: deep_river
-Version: 0.3.2
+Version: 0.4.0
 Summary: Online Deep Learning for river
 Author-email: Cedric Kulbach <cedric.kulbach@googlemail.com>, Lucas Cazzonelli <Cazzonelli@fzi.de>, Hoang-Anh Ngo <hoang-anh.ngo@alumni.polytechnique.org>, Max Halford <maxhalford25@gmail.com>, Saulo Martiello Mastelini <saulomastelini@gmail.com>
 License: BSD-3-Clause
@@ -8,71 +8,31 @@ Project-URL: Homepage, https://github.com/online-ml/deep-river
 Project-URL: Repository, https://github.com/online-ml/deep-river
 Project-URL: Documentation, https://github.com/online-ml/deep-river
 Keywords: online-learning,deep-learning,data-streams
-Classifier: Programming Language :: Python
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Classifier: Programming Language :: Python :: Implementation :: CPython
-Classifier: Programming Language :: Python :: Implementation :: PyPy
-Requires-Python: <3.13,>=3.10
+Requires-Python: <3.13,>=3.11
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: scikit-learn~=1.5.0
-Requires-Dist: torch==2.2.2
-Requires-Dist: pandas~=2.2.2
-Requires-Dist: numpy~=1.26.4
-Requires-Dist: river
+Requires-Dist: numpy<3,>=2.3.4
+Requires-Dist: pandas<3,>=2.2
+Requires-Dist: river<0.27,>=0.26
+Requires-Dist: scikit-learn<2,>=1.5
+Requires-Dist: sortedcontainers>=2.4.0
+Requires-Dist: torch==2.2.2; sys_platform == "darwin" and platform_machine == "x86_64"
+Requires-Dist: torch<2.7,>=2.5; sys_platform != "darwin" or platform_machine != "x86_64"
 Requires-Dist: tqdm~=4.66.5
 Requires-Dist: torchviz~=0.0.2
-Requires-Dist: mkdocstrings[python]>=0.19.0
-Requires-Dist: pytkdocs[numpy-style]>=0.5.0
-Requires-Dist: sortedcontainers>=2.4.0
-Provides-Extra: dev
-Requires-Dist: graphviz>=0.20.3; extra == "dev"
-Requires-Dist: matplotlib>=3.9.2; extra == "dev"
-Requires-Dist: mypy>=1.11.1; extra == "dev"
-Requires-Dist: codecov>=2.1.13; extra == "dev"
-Requires-Dist: pre-commit>=3.8.0; extra == "dev"
-Requires-Dist: pytest>=8.3.2; extra == "dev"
-Requires-Dist: pytest-cov>=5.0.0; extra == "dev"
-Requires-Dist: black>=24.8.0; extra == "dev"
-Requires-Dist: flake8>=7.1.1; extra == "dev"
-Requires-Dist: isort>=5.13.2; extra == "dev"
-Requires-Dist: jupyter>=1.0.0; extra == "dev"
-Requires-Dist: pyupgrade==3.17.0; extra == "dev"
-Requires-Dist: flask>=3.0.2; extra == "dev"
-Requires-Dist: ipykernel>=6.9.0; extra == "dev"
-Requires-Dist: mike>=0.5.3; extra == "dev"
-Requires-Dist: mkdocs>=1.2.3; extra == "dev"
-Requires-Dist: mkdocs-awesome-pages-plugin>=2.7.0; extra == "dev"
-Requires-Dist: mkdocs-gen-files>=0.3.5; extra == "dev"
-Requires-Dist: mkdocs-charts-plugin>=0.0.8; extra == "dev"
-Requires-Dist: mkdocs-literate-nav>=0.4.1; extra == "dev"
-Requires-Dist: mkdocs-material>=8.1.11; extra == "dev"
-Requires-Dist: ipython_genutils>=0.1.0; extra == "dev"
-Requires-Dist: mkdocs-jupyter>=0.20.0; extra == "dev"
-Requires-Dist: nbconvert>=6.4.2; extra == "dev"
-Requires-Dist: jupyter_contrib_nbextensions==0.7.0; extra == "dev"
-Requires-Dist: numpydoc>=1.2; extra == "dev"
-Requires-Dist: notebook==6.4.3; extra == "dev"
-Requires-Dist: spacy>=3.2.2; extra == "dev"
-Requires-Dist: jinja2>=3.0.3; extra == "dev"
-Requires-Dist: dominate; extra == "dev"
-Requires-Dist: jupyter-client; extra == "dev"
-Requires-Dist: python-slugify; extra == "dev"
-Requires-Dist: watermark==2.3.1; extra == "dev"
-Requires-Dist: plotly>=5.0.0; extra == "dev"
 Dynamic: license-file
 
 <p align="center">
-  <img height="150px" src="https://raw.githubusercontent.com/online-ml/deep-river/master/docs/img/logo.png" alt="incremental dl logo">
+  <img height="150px" src="https://raw.githubusercontent.com/online-ml/deep-river/main/docs/img/logo.png" alt="incremental dl logo">
 </p>
 <p align="center">
     <img alt="PyPI" src="https://img.shields.io/pypi/v/deep-river">
     <!-- Single Coverage Badge (Shields.io) -->
     <a href="https://app.codecov.io/gh/online-ml/deep-river" >
-        <img alt="Coverage" src="https://img.shields.io/codecov/c/github/online-ml/deep-river?branch=master&logo=codecov&label=coverage" />
+        <img alt="Coverage" src="https://img.shields.io/codecov/c/github/online-ml/deep-river?branch=main&logo=codecov&label=coverage" />
     </a>
     <img alt="PyPI - Python Version" src="https://img.shields.io/pypi/pyversions/deep-river">
     <img alt="PyPI - Downloads" src="https://img.shields.io/pypi/dm/deep-river">
@@ -88,7 +48,7 @@ Dynamic: license-file
 
 ## 📚 [Documentation](https://online-ml.github.io/deep-river/)
 The [documentation](https://online-ml.github.io/deep-river/) contains an overview of all features of this repository as well as the repository's full features list. 
-In each of these, the git repo reference is listed in a section that shows [examples](https://github.com/online-ml/deep-river/blob/master/docs/examples) of the features and functionality.
+In each of these, the git repo reference is listed in a section that shows [examples](https://github.com/online-ml/deep-river/blob/main/docs/examples) of the features and functionality.
 As we are always looking for further use cases and examples, feel free to contribute to the documentation or the repository itself via a pull request
 
 ## 💈 Installation
@@ -103,7 +63,7 @@ pip install "river[deep]"
 You can install the latest development version from GitHub as so:
 
 ```shell
-pip install https://github.com/online-ml/deep-river/archive/refs/heads/master.zip
+pip install https://github.com/online-ml/deep-river/archive/refs/heads/main.zip
 ```
 
 ### Development Environment
@@ -119,7 +79,7 @@ git clone https://github.com/online-ml/deep-river.git
 cd deep-river
 
 # Install all dependencies (including dev dependencies)
-uv sync --extra dev
+uv sync --group dev
 
 # Run tests
 make test
@@ -316,7 +276,7 @@ To acknowledge the use of the `DeepRiver` library in your research, please refer
 ## 🏫 Affiliations
 
 <p align="center">
-    <img src="https://upload.wikimedia.org/wikipedia/de/thumb/4/44/Fzi_logo.svg/1200px-Fzi_logo.svg.png?raw=true" alt="FZI Logo" height="200"/>
+    <img src="https://upload.wikimedia.org/wikipedia/de/4/44/Fzi_logo.svg" alt="FZI Logo" height="200"/>
 </p>
 
 <p align="center">

{deep_river-0.3.2 → deep_river-0.4.0}/README.md

@@ -1,11 +1,11 @@
 <p align="center">
-  <img height="150px" src="https://raw.githubusercontent.com/online-ml/deep-river/master/docs/img/logo.png" alt="incremental dl logo">
+  <img height="150px" src="https://raw.githubusercontent.com/online-ml/deep-river/main/docs/img/logo.png" alt="incremental dl logo">
 </p>
 <p align="center">
     <img alt="PyPI" src="https://img.shields.io/pypi/v/deep-river">
     <!-- Single Coverage Badge (Shields.io) -->
     <a href="https://app.codecov.io/gh/online-ml/deep-river" >
-        <img alt="Coverage" src="https://img.shields.io/codecov/c/github/online-ml/deep-river?branch=master&logo=codecov&label=coverage" />
+        <img alt="Coverage" src="https://img.shields.io/codecov/c/github/online-ml/deep-river?branch=main&logo=codecov&label=coverage" />
     </a>
     <img alt="PyPI - Python Version" src="https://img.shields.io/pypi/pyversions/deep-river">
     <img alt="PyPI - Downloads" src="https://img.shields.io/pypi/dm/deep-river">
@@ -21,7 +21,7 @@
 
 ## 📚 [Documentation](https://online-ml.github.io/deep-river/)
 The [documentation](https://online-ml.github.io/deep-river/) contains an overview of all features of this repository as well as the repository's full features list. 
-In each of these, the git repo reference is listed in a section that shows [examples](https://github.com/online-ml/deep-river/blob/master/docs/examples) of the features and functionality.
+In each of these, the git repo reference is listed in a section that shows [examples](https://github.com/online-ml/deep-river/blob/main/docs/examples) of the features and functionality.
 As we are always looking for further use cases and examples, feel free to contribute to the documentation or the repository itself via a pull request
 
 ## 💈 Installation
@@ -36,7 +36,7 @@ pip install "river[deep]"
 You can install the latest development version from GitHub as so:
 
 ```shell
-pip install https://github.com/online-ml/deep-river/archive/refs/heads/master.zip
+pip install https://github.com/online-ml/deep-river/archive/refs/heads/main.zip
 ```
 
 ### Development Environment
@@ -52,7 +52,7 @@ git clone https://github.com/online-ml/deep-river.git
 cd deep-river
 
 # Install all dependencies (including dev dependencies)
-uv sync --extra dev
+uv sync --group dev
 
 # Run tests
 make test
@@ -249,7 +249,7 @@ To acknowledge the use of the `DeepRiver` library in your research, please refer
 ## 🏫 Affiliations
 
 <p align="center">
-    <img src="https://upload.wikimedia.org/wikipedia/de/thumb/4/44/Fzi_logo.svg/1200px-Fzi_logo.svg.png?raw=true" alt="FZI Logo" height="200"/>
+    <img src="https://upload.wikimedia.org/wikipedia/de/4/44/Fzi_logo.svg" alt="FZI Logo" height="200"/>
 </p>
 
 <p align="center">

deep_river-0.4.0/deep_river/__init__.py

@@ -0,0 +1,7 @@
+from importlib.metadata import version as _version
+
+from . import anomaly, classification, forecasting, regression, utils
+
+__version__ = _version("deep-river")
+
+__all__ = ["anomaly", "classification", "forecasting", "regression", "utils"]

{deep_river-0.3.2 → deep_river-0.4.0}/deep_river/anomaly/__init__.py

@@ -1,12 +1,13 @@
+"""
+This module contains the anomaly detection algorithms for the
+deep_river package.
+"""
+
 from .ae import Autoencoder
 from .probability_weighted_ae import ProbabilityWeightedAutoencoder
 from .rolling_ae import RollingAutoencoder
 from .scaler import AnomalyMeanScaler, AnomalyMinMaxScaler, AnomalyStandardScaler
 
-"""
-This module contains the anomaly detection algorithms for the
-deep_river package.
-"""
 __all__ = [
     "AnomalyStandardScaler",
     "AnomalyMeanScaler",

{deep_river-0.3.2 → deep_river-0.4.0}/deep_river/anomaly/ae.py

@@ -3,7 +3,7 @@ from typing import Any, Callable, Union
 import numpy as np
 import pandas as pd
 import torch
-from river.anomaly.base import AnomalyDetector
+from river.base import AnomalyDetector
 from torch import nn
 
 from deep_river.base import DeepEstimator
@@ -108,7 +108,17 @@ class Autoencoder(DeepEstimator, AnomalyDetector):
         """
         return set()
 
-    def learn_one(self, x: dict, y: Any = None, **kwargs) -> None:
+    def _update_observed_features(self, x: dict | pd.DataFrame) -> bool:
+        updated = super()._update_observed_features(x)
+        if (
+            self.is_feature_incremental
+            and self.output_layer is not None
+            and self._get_output_size() < self._get_input_size()
+        ):
+            self._expand_layer(self.output_layer, self._get_input_size(), output=True)
+        return updated
+
+    def learn_one(self, x: dict, y: Any = None) -> None:
         """
         Performs one step of training with a single example.
 
@@ -116,8 +126,6 @@ class Autoencoder(DeepEstimator, AnomalyDetector):
         ----------
         x
             Input example.
-
-        **kwargs
         """
         self._update_observed_features(x)
         self._learn(self._dict2tensor(x))
@@ -188,5 +196,5 @@ class Autoencoder(DeepEstimator, AnomalyDetector):
             self.loss_func(x_pred, x_t, reduction="none"),
             dim=list(range(1, x_t.dim())),
         )
-        score = loss.cpu().detach().numpy()
+        score = np.asarray(loss.cpu().detach().tolist())
         return score

{deep_river-0.3.2 → deep_river-0.4.0}/deep_river/anomaly/probability_weighted_ae.py

@@ -1,6 +1,7 @@
 import math
 from typing import Any, Callable, Union
 
+import numpy as np
 import pandas as pd
 import torch
 from river import stats, utils
@@ -21,9 +22,13 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
         device: str = "cpu",
         seed: int = 42,
         skip_threshold: float = 0.9,
-        window_size=250,
+        window_size: int = 250,
         **kwargs,
     ):
+        if not 0 < skip_threshold <= 1:
+            raise ValueError("skip_threshold must be greater than 0 and at most 1.")
+        if window_size < 1:
+            raise ValueError("window_size must be positive.")
         super().__init__(
             module=module,
             loss_fn=loss_fn,
@@ -38,7 +43,7 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
         self.rolling_mean = utils.Rolling(stats.Mean(), window_size=window_size)
         self.rolling_var = utils.Rolling(stats.Var(), window_size=window_size)
 
-    def learn_one(self, x: dict, y: Any = None, **kwargs) -> None:
+    def learn_one(self, x: dict, y: Any = None) -> None:
         """
         Performs one step of training with a single example,
         scaling the employed learning rate based on the outlier
@@ -46,7 +51,6 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
 
         Parameters
         ----------
-        **kwargs
         x
             Input example.
 
@@ -65,23 +69,29 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
         self._apply_loss(loss)
 
     def _apply_loss(self, loss):
-        losses_numpy = loss.detach().numpy()
+        losses_numpy = np.asarray(loss.detach().cpu().tolist())
         mean = self.rolling_mean.get()
         var = self.rolling_var.get() if self.rolling_var.get() > 0 else 1
-        if losses_numpy.ndim == 0:
-            self.rolling_mean.update(losses_numpy)
-            self.rolling_var.update(losses_numpy)
-        else:
-            for loss_numpy in range(len(losses_numpy)):
-                self.rolling_mean.update(loss_numpy)
-                self.rolling_var.update(loss_numpy)
+        for loss_numpy in np.atleast_1d(losses_numpy):
+            self.rolling_mean.update(float(loss_numpy))
+            self.rolling_var.update(float(loss_numpy))
 
         loss_scaled = (losses_numpy - mean) / math.sqrt(var)
         prob = ndtr(loss_scaled)
-        loss = torch.tensor((self.skip_threshold - prob) / self.skip_threshold) * loss
-
+        weights = loss.new_tensor(
+            np.maximum(0, (self.skip_threshold - prob) / self.skip_threshold).tolist()
+        )
         self.optimizer.zero_grad()
+        if not weights.numel():
+            return
+        loss = (weights * loss).mean()
         loss.backward()
+        if not torch.any(weights > 0):
+            return
+        if self.gradient_clip_value is not None:
+            torch.nn.utils.clip_grad_norm_(
+                self.module.parameters(), self.gradient_clip_value
+            )
         self.optimizer.step()
 
     def learn_many(self, X: pd.DataFrame) -> None:
@@ -89,7 +99,7 @@ class ProbabilityWeightedAutoencoder(ae.Autoencoder):
         X_t = self._df2tensor(X)
 
         self.module.train()
-        x_pred = self.module(X)
+        x_pred = self.module(X_t)
         loss = torch.mean(
             self.loss_func(x_pred, X_t, reduction="none"),
             dim=list(range(1, X_t.dim())),

{deep_river-0.3.2 → deep_river-0.4.0}/deep_river/anomaly/rolling_ae.py

@@ -1,9 +1,8 @@
 from typing import Any, Callable, List, Union
 
-import numpy as np
 import pandas as pd
 import torch
-from river import anomaly
+from river.base import AnomalyDetector
 from torch import nn
 
 from deep_river.base import RollingDeepEstimator
@@ -55,7 +54,7 @@ class _TestLSTMAutoencoder(nn.Module):
         return output
 
 
-class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
+class RollingAutoencoder(RollingDeepEstimator, AnomalyDetector):
     """Rolling window autoencoder for streaming anomaly detection.
 
     Maintains a fixed-size deque of the latest ``window_size`` observations and
@@ -141,7 +140,17 @@ class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
             "check_predict_proba_one_binary",
         }
 
-    def learn_one(self, x: dict, y: Any = None, **kwargs) -> None:
+    def _update_observed_features(self, x: dict | pd.DataFrame) -> bool:
+        updated = super()._update_observed_features(x)
+        if (
+            self.is_feature_incremental
+            and self.output_layer is not None
+            and self._get_output_size() < self._get_input_size()
+        ):
+            self._expand_layer(self.output_layer, self._get_input_size(), output=True)
+        return updated
+
+    def learn_one(self, x: dict, y: Any = None) -> None:
         """Update model using a single sample appended to the rolling window.
 
         Parameters
@@ -150,31 +159,17 @@ class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
             Dictionary containing feature name-value pairs for the sample.
         y : Any, optional
             Target value (not used in autoencoder training).
-        **kwargs
-            Additional keyword arguments.
         """
         self._update_observed_features(x)
-        self._x_window.append(list(x.values()))
+        self._x_window.append([x.get(feature, 0) for feature in self.observed_features])
 
         x_t = deque2rolling_tensor(self._x_window, device=self.device)
         self._learn(x=x_t)
 
     def learn_many(self, X: pd.DataFrame, y=None) -> None:
-        """Batch update; extends window with rows from X and learns if full.
-
-        Parameters
-        ----------
-        X : pd.DataFrame
-            DataFrame containing the input features for each sample.
-        y : None
-            Ignored, present for compatibility.
-        """
-        self._update_observed_features(X)
-
-        self._x_window.append(X.values.tolist())
-        if len(self._x_window) == self.window_size:
-            X_t = deque2rolling_tensor(self._x_window, device=self.device)
-            self._learn(x=X_t)
+        columns = X.columns
+        for values in X.to_numpy(dtype=object).tolist():
+            self.learn_one(dict(zip(columns, values)))
 
     def score_one(self, x: dict) -> float:
         """Return reconstruction error for current window + candidate sample.
@@ -193,7 +188,7 @@ class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
         self._update_observed_features(x)
         if len(self._x_window) == self.window_size:
             x_win = self._x_window.copy()
-            x_win.append(list(x.values()))
+            x_win.append([x.get(feature, 0) for feature in self.observed_features])
             x_t = deque2rolling_tensor(x_win, device=self.device)
             self.module.eval()
             with torch.inference_mode():
@@ -202,7 +197,9 @@ class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
             res = loss.item()
 
         if self.append_predict:
-            self._x_window.append(list(x.values()))
+            self._x_window.append(
+                [x.get(feature, 0) for feature in self.observed_features]
+            )
         return res
 
     def score_many(self, X: pd.DataFrame) -> List[Any]:
@@ -220,24 +217,4 @@ class RollingAutoencoder(RollingDeepEstimator, anomaly.base.AnomalyDetector):
         List[float]
             List of computed anomaly scores (reconstruction errors) for each sample in X.
         """
-        self._update_observed_features(X)
-        x_win = self._x_window.copy()
-        x_win.append(X.values.tolist())
-        if self.append_predict:
-            self._x_window.append(X.values.tolist())
-
-        if len(self._x_window) == self.window_size:
-            X_t = deque2rolling_tensor(x_win, device=self.device)
-            self.module.eval()
-            with torch.inference_mode():
-                x_pred = self.module(X_t)
-            loss = torch.mean(
-                self.loss_func(x_pred, x_pred, reduction="none"),
-                dim=list(range(1, x_pred.dim())),
-            )
-            losses = loss.detach().numpy()
-            if len(losses) < len(X):
-                losses = np.pad(losses, (len(X) - len(losses), 0))
-            return losses.tolist()
-        else:
-            return np.zeros(len(X)).tolist()
+        return [self.score_one(row) for row in X.to_dict(orient="records")]

{deep_river-0.3.2 → deep_river-0.4.0}/deep_river/anomaly/scaler.py

@@ -1,9 +1,10 @@
 import abc
 
 import numpy as np
+import pandas as pd
 from river import base, utils
 from river.anomaly import HalfSpaceTrees
-from river.anomaly.base import AnomalyDetector
+from river.base import AnomalyDetector
 from river.stats import Mean, Min, RollingMin
 
 
@@ -21,7 +22,7 @@ class AnomalyScaler(base.Wrapper, AnomalyDetector):
         self.anomaly_detector = anomaly_detector
 
     @classmethod
-    def _unit_test_params(self) -> dict:
+    def _unit_test_params(cls):
         """
         Returns a dictionary of parameters to be used for unit testing
         the respective class.
@@ -32,7 +33,7 @@ class AnomalyScaler(base.Wrapper, AnomalyDetector):
             Dictionary of parameters to be used for unit testing the
             respective class.
         """
-        return {"anomaly_detector": HalfSpaceTrees()}
+        yield {"anomaly_detector": HalfSpaceTrees(seed=42, window_size=10)}
 
     @classmethod
     def _unit_test_skips(self) -> set:
@@ -60,7 +61,7 @@ class AnomalyScaler(base.Wrapper, AnomalyDetector):
         return self.anomaly_detector
 
     @abc.abstractmethod
-    def score_one(self, *args) -> float:
+    def score_one(self, *args, **kwargs) -> float:
         """Return a scaled anomaly score based on raw score provided by
         the wrapped anomaly detector.
 
@@ -79,7 +80,7 @@ class AnomalyScaler(base.Wrapper, AnomalyDetector):
         more anomalous examples.
         """
 
-    def learn_one(self, *args) -> None:
+    def learn_one(self, *args, **kwargs) -> None:
         """
         Update the scaler and the underlying anomaly scaler.
 
@@ -95,26 +96,22 @@ class AnomalyScaler(base.Wrapper, AnomalyDetector):
             The model itself.
         """
 
-        self.anomaly_detector.learn_one(*args)
+        self._update_score(self.anomaly_detector.score_one(*args, **kwargs))
+        self.anomaly_detector.learn_one(*args, **kwargs)
 
     @abc.abstractmethod
-    def score_many(self, *args) -> np.ndarray:
-        """Return scaled anomaly scores based on raw score provided by
-        the wrapped anomaly detector.
+    def _update_score(self, score: float) -> None:
+        pass
 
-        A high score is indicative of an anomaly. A low score corresponds
-        to a normal observation.
-
-        Parameters
-        ----------
-        *args
-            Depends on whether the underlying anomaly detector is
-            supervised or not.
+    def score_many(self, X: pd.DataFrame, **kwargs) -> np.ndarray:
+        return np.asarray(
+            [self.score_one(x, **kwargs) for x in X.to_dict(orient="records")],
+            dtype=float,
+        )
 
-        Returns
-        -------
-        Scaled anomaly scores. Larger values indicate more anomalous examples.
-        """
+    def learn_many(self, X: pd.DataFrame, **kwargs) -> None:
+        for x in X.to_dict(orient="records"):
+            self.learn_one(x, **kwargs)
 
 
 class AnomalyStandardScaler(AnomalyScaler):
@@ -141,16 +138,20 @@ class AnomalyStandardScaler(AnomalyScaler):
         rolling: bool = True,
         window_size: int = 250,
     ):
+        from river.stats import Var
+
         super().__init__(anomaly_detector)
         self.rolling = rolling
         self.window_size = window_size
         self.mean = utils.Rolling(Mean(), self.window_size) if self.rolling else Mean()
-        self.sq_mean = (
-            utils.Rolling(Mean(), self.window_size) if self.rolling else Mean()
+        self.var = (
+            utils.Rolling(Var(ddof=0), self.window_size)
+            if self.rolling
+            else Var(ddof=0)
         )
         self.with_std = with_std
 
-    def score_one(self, *args):
+    def score_one(self, *args, **kwargs):
         """
         Return a scaled anomaly score based on raw score provided by the
         wrapped anomaly detector. Larger values indicate more
@@ -167,17 +168,16 @@ class AnomalyStandardScaler(AnomalyScaler):
         An scaled anomaly score. Larger values indicate more
         anomalous examples.
         """
-        raw_score = self.anomaly_detector.score_one(*args)
-        mean = self.mean.update(raw_score).get()
-        if self.with_std:
-            var = (
-                self.sq_mean.update(raw_score**2).get() - mean**2
-            )  # todo is this correct?
-            score = (raw_score - mean) / var**0.5
-        else:
-            score = raw_score - mean
+        raw_score = self.anomaly_detector.score_one(*args, **kwargs)
+        mean = self.mean.get()
+        if not self.with_std:
+            return raw_score - mean
+        var = self.var.get()
+        return (raw_score - mean) / var**0.5 if var > 0 else 0.0
 
-        return score
+    def _update_score(self, score: float) -> None:
+        self.mean.update(score)
+        self.var.update(score)
 
 
 class AnomalyMeanScaler(AnomalyScaler):
@@ -200,14 +200,14 @@ class AnomalyMeanScaler(AnomalyScaler):
         self,
         anomaly_detector: AnomalyDetector,
         rolling: bool = True,
-        window_size=250,
+        window_size: int = 250,
     ):
         super().__init__(anomaly_detector=anomaly_detector)
         self.rolling = rolling
         self.window_size = window_size
         self.mean = utils.Rolling(Mean(), self.window_size) if self.rolling else Mean()
 
-    def score_one(self, *args):
+    def score_one(self, *args, **kwargs):
         """
         Return a scaled anomaly score based on raw score provided by the
         wrapped anomaly detector. Larger values indicate more
@@ -224,11 +224,12 @@ class AnomalyMeanScaler(AnomalyScaler):
         An scaled anomaly score. Larger values indicate more
         anomalous examples.
         """
-        raw_score = self.anomaly_detector.score_one(*args)
-        mean = self.mean.update(raw_score).get()
-        score = raw_score / mean
+        raw_score = self.anomaly_detector.score_one(*args, **kwargs)
+        mean = self.mean.get()
+        return raw_score / mean if mean else 0.0
 
-        return score
+    def _update_score(self, score: float) -> None:
+        self.mean.update(score)
 
 
 class AnomalyMinMaxScaler(AnomalyScaler):
@@ -251,13 +252,15 @@ class AnomalyMinMaxScaler(AnomalyScaler):
         rolling: bool = True,
         window_size: int = 250,
     ):
+        from river.stats import Max, RollingMax
+
         super().__init__(anomaly_detector)
         self.rolling = rolling
         self.window_size = window_size
         self.min = RollingMin(self.window_size) if self.rolling else Min()
-        self.max = RollingMin(self.window_size) if self.rolling else Min()
+        self.max = RollingMax(self.window_size) if self.rolling else Max()
 
-    def score_one(self, *args):
+    def score_one(self, *args, **kwargs):
         """
         Return a scaled anomaly score based on raw score provided by the
         wrapped anomaly detector. Larger values indicate more
@@ -274,9 +277,13 @@ class AnomalyMinMaxScaler(AnomalyScaler):
         An scaled anomaly score. Larger values indicate more
         anomalous examples.
         """
-        raw_score = self.anomaly_detector.score_one(*args)
-        min = self.min.update(raw_score).get()
-        max = self.max.update(raw_score).get()
-        score = (raw_score - min) / (max - min)
-
-        return score
+        raw_score = self.anomaly_detector.score_one(*args, **kwargs)
+        minimum = self.min.get()
+        maximum = self.max.get()
+        if minimum is None or maximum is None:
+            return 0.0
+        return (raw_score - minimum) / (maximum - minimum) if maximum > minimum else 0.0
+
+    def _update_score(self, score: float) -> None:
+        self.min.update(score)
+        self.max.update(score)